基于ICA方法和遗传算法的胎儿心电信号提取

FECG Extraction Based on ICA and Genetic Algorithm

现有独立分量分析(ICA)方法易陷入局部极优值,从而使提取出的胎儿心电信号(FECG)中常混有部分母体心电(MECG)。为解决该问题,本文提出一种基于独立分量分析和遗传算法(GA)的新的胎儿心电信号提取方法。该方法首先利用遗传算法在全局空间内搜索最优值,然后又通过基于峭度的固定点算法的迭代核加强局部搜索,因而增强了算法的分离能力。该算法适用于超高斯和亚高斯信号同时存在的情况,并且具有遗传算法所有的收敛性强的优点。实验结果表明,该算法可以成功地分离混叠信号,而且与传统的独立分量分析方法相比,具有更优异的分离能力,提取出的胎儿心电信号噪声小,几乎不混有母体心电。

The existing independent component analysis（ICA） methods are easy to strap into local optimum values,so the extracted fetal electrocardiogram（FECG） is often mixed with some maternal ECG（MECG）.To solve this problem,a new fetal ECG extraction method is proposed based on the independent component analysis and the genetic algorithm（GA）.Firstly,this method searches the optimum value through the whole space using GA,and uses the iteration core of the fixedpoint algorithm based on the kurtosis to strengthen the local search,thus enhancing the capacity of the separation.The method can be used under super-Gaussian and sub-Gaussian mixed conditions.In addition,it has the advantage of strong convergence for GA.Simulation result shows that the method can separate the mixed signals,and it is superior to traditional ICA methods in separating the mixed signals.And the method can extract fetal ECG with less noise,and there is almost no maternal ECG in FECG.